FLUX 3 Max "Ties for First" at 22 Times Seedance's Speed. The Tie Is an Overlapping Confidence Interval, and the Price Quadruples After November 7.
fal launched FLUX 3 Max on October 19, a video model post-trained on Black Forest Labs' FLUX 3 Video, with vendor timings about 22 times faster than Seedance 2.5 and a claim of tied-for-first quality. Artificial Analysis's image-to-video board puts it two Elo behind MiniMax H3 with overlapping confidence intervals and a rank range of first to fifth. The 75% launch discount, $0.0425 per second at 720p, ends November 7; list is $0.17 per second. API only, no weights. We read the press release, VentureBeat's read of AA's table, and fal's pricing.
By T.W. Ghost
When we wrote about Seedance 2.5 in August, the point was the gap between a video model's marketing and the settings you can actually run. Today's launch has the same shape. fal released FLUX 3 Max on October 19, 2026, a video model post-trained on Black Forest Labs' FLUX 3 Video, with two claims: tied-for-first quality on Artificial Analysis's image-to-video board, and roughly 22 times Seedance 2.5's speed.
Both claims have a source, and both sources say something narrower than the headline. The quality tie is an overlapping confidence interval on a board where FLUX 3 Max's rank range runs from first to fifth. The speed is fal's own stopwatch, and fal's two announcements disagree with each other by a second and a half. And the price fal is advertising ends on November 7, after which it quadruples. We read the press release, VentureBeat's October 19 read of AA's table, and fal's pricing. Here is what held up.
What You Are Buying
| FLUX 3 Max (fal) | |
|---|---|
| Launched | October 19, 2026; API only; no weights |
| Base | Post-trained on Black Forest Labs' FLUX 3 Video |
| Inputs | Text, image, or up to 10 keyframes |
| Output | 5 to 20 seconds; 720p native; optional 1080p upscale; optional audio |
| Promo price (through Nov 7) | $0.0425 per second at 720p; $0.0725 at 1080p |
| List price (from Nov 8) | $0.17 per second at 720p; $0.29 at 1080p |
| Vendor speed claim | 10-second clip in 12.45 s average vs 277.80 s for Seedance 2.5 (VentureBeat's figures from fal); the press release says 10.95 s |
| AA image-to-video (as of Oct 19) | 1,138 Elo (CI 1,123 to 1,153); MiniMax H3 1,140 (1,124 to 1,156); Seedance 2.5 1,138 (1,121 to 1,155); FLUX rank range 1 to 5 |
The tie is a statistical tie, which is to say not a win. On the table VentureBeat read from AA on October 19, MiniMax H3 is first by two Elo, with all three confidence intervals overlapping and FLUX 3 Max's rank anywhere from first to fifth. fal's launch materials used a preliminary snapshot of about 1,139. "In the H3 and Seedance cluster" is the accurate sentence.
The stopwatch is fal's. 12.45 seconds against 277.80 for Seedance 2.5 on a 10-second 720p clip is fal's measurement as relayed by VentureBeat; fal's own press release prints 10.95 seconds for the same comparison. We print both and call it about 11 to 12.5 seconds. fal's docs also warn that settings change latency, so do not promise the number to your users.
The price cliff is the real catch. By our arithmetic, a 10-second 720p clip costs $0.425 at the promo rate and $1.70 at list. A ten-attempt creative loop goes from $4.25 to $17 the morning after November 7. AA's API pricing column reportedly shows $2.55 per minute for FLUX 3 Max at 720p; that is the promo rate times 60, so it will move too.
The Same Lesson as Seedance
The pattern from the Seedance post holds: the leaderboard measures quality with audio on image-to-video, the launch quotes it as a leadership claim, and the settings you can afford at scale are a different question. Three honest observations:
- I2V is not T2V. The AA figures fal cites are for image-to-video with audio. Text-to-video standing is not established by them.
- No weights, no self-hosting. Unlike BFL's own FLUX 3 Image Klein from four days earlier, FLUX 3 Max is API only, with commercial use of outputs under fal's terms.
- The board moves. AA's Elo and intervals are a snapshot from October 19; re-check the morning you cite them.
Verified vs Unconfirmed: The Scorecard
| Claim | Verdict |
|---|---|
| Launched Oct 19; post-trained on FLUX 3 Video; 5 to 20 s; 720p native; 1080p upscale; up to 10 keyframes; API only | Vendor and press (PR Newswire, VentureBeat, fal docs) |
| Tied for first on AA image-to-video | Contradicted as stated: second to H3 by 2 Elo with overlapping intervals; rank range 1 to 5 (VentureBeat reading AA) |
| About 22x Seedance 2.5's speed | Vendor (fal's timing); the two fal figures disagree (10.95 s vs 12.45 s) |
| Promo $0.0425 / $0.0725 per second through Nov 7; list $0.17 / $0.29 | Press (VentureBeat's table from fal) |
| 10 s at promo = $0.425; at list = $1.70 | Derived (our arithmetic) |
| AA $2.55 per minute column | Secondary; consistent with promo math |
| Text-to-video ranking | Unverified |
Who Should Use It
Use FLUX 3 Max before November 7 for image-to-video work where a fast, cheap, keyframe-driven clip in the H3 and Seedance quality cluster is what you need, and budget every production estimate at the list price of $0.17 per second unless you are certain you finish first.
Skip it if your pipeline needs weights, text-to-video leadership, or a wall-clock promise; fal's own docs will not give you the last one.
The frontier is elsewhere: on AA's image-to-video board the top three are within a few Elo of each other and the intervals overlap. Whoever is first this morning is a coin flip by the methodology's own error bars. Our model comparison has the current standings side by side.
Sources
- VentureBeat: fal launches FLUX 3 Max
- PR Newswire: fal launches FLUX 3 Max
- Artificial Analysis: image-to-video board
*Not sure whether speed, quality, or the price after the promo matters most for your video pipeline? Take the free 2-minute quiz and get matched. Then read the August post this one rhymes with: Seedance 2.5 and the 4K claim.*